Abstract
OBJECTIVES: This study aimed to establish a novel and practical nomogram for use upon hospital admission to identify coinfections among elderly patients with coronavirus disease 2019 (COVID-19) to provide timely intervention, limit antimicrobial agent overuse, and finally reduce unfavourable outcomes. METHODS: This prospective cohort study included COVID-19 patients consecutively admitted at multicenter medical facilities in a two-stage process. The nomogram was built on the multivariable logistic regression analysis. The performance of the nomogram was assessed for discrimination and calibration using receiver operating characteristic curves, calibration plots, and decision curve analysis (DCA) in rigorous internal and external validation settings. Two different cutoff values were determined to stratify coinfection risk in elderly patients with COVID-19. RESULTS: The coinfection rates in elderly patients determined to be and 26.61%. The nomogram was developed with the parameters of diabetes comorbidity, previous invasive procedure, and procalcitonin (PCT) level, which together showed areas under the curve of 0.86, 0.82, and 0.83 in the training, internal validation, and external validation cohorts, respectively. The nomogram outperformed both PCT or C-reactive protein level alone in detecting coinfections in elderly patients with COVID-19; in addition, we found the nomogram was specific for the elderly compared to non-elderly group. To facilitate clinical decision-making among elderly patients with COVID-19, we defined two cutoff values of prediction probability: a low cutoff of 6.65% to rule out coinfections and a high cutoff of 27.79% to confidently confirm coinfections. CONCLUSIONS: This novel nomogram will assist in the early identification of coinfections in elderly patients with COVID-19.